The Association of Vision, Hearing, and Dual-Sensory Loss with Walking Speed and Incident Slow Walking: Longitudinal and Time to Event Analyses in the Health and Retirement Study

Abstract

With the aging of the population, vision (VL), hearing (HL), and dual-sensory (DSL, concurrent VL and HL) loss will likely constitute important public health challenges. Walking speed is an indicator of functional status and is associated with mortality. Using the Health and Retirement Study, a nationally representative U.S. cohort, we analyzed the longitudinal relationship between sensory loss and walking speed. In multivariable mixed effects linear models, baseline walking speed was slower by 0.05 m/s (95% confidence interval [CI] = 0.04–0.07) for VL, 0.02 (95% CI = 0.003–0.03) for HL, and 0.07 (95% CI = 0.05–0.08) for DSL compared with those without sensory loss. Similar annual declines in walking speeds occurred in all groups. In time-to-event analyses, the risk of incident slow walking speed (walking speed < 0.6 m/s) was 43% (95% CI = 25–65%), 29% (95% CI = 13–48%), and 35% (95% CI = 13–61%) higher among those with VL, HL, and DSL respectively, relative to those without sensory loss. The risk of incident very slow walking speed (walking speed < 0.4 m/s) was significantly higher among those with HL and DSL relative to those without sensory loss, and significantly higher among those with DSL relative to those with VL or HL alone. Addressing sensory loss and teaching compensatory strategies may help mitigate the effect of sensory loss on walking speed.

By 2030, 20% of the U.S. population will be 65 years or older. ¹ Providing medical care to older adults can be especially challenging, as the goals and plans of care may differ significantly from patient to patient based on functional status and life expectancy. Walking speed, or gait speed, is an established indicator of functional status and has been associated with survival. ² Thus, it is an important measure when evaluating older adults, and has been recognized by some as the sixth vital sign of aging. ^{2 3}

Over 80% of adults 80 years and older have hearing (HL) and/or vision loss (VL). Both HL and VL will likely constitute important public health challenges with the anticipated aging of the population. ^{1 4 5} Dual-sensory loss (DSL; concurrent VL and HL) affects approximately one in nine adults older than 80 years, ⁶ and will similarly affect a growing number of people. DSL has been linked to adverse health, ^{7 8 9 10} communication problems, ¹¹ reduced quality of life, ¹² and increased mortality. ^{13 14} Across these adverse outcomes, DSL is typically associated with a worse prognosis than either sensory loss alone.

Poor vision and hearing loss are linked to reduced subjectively reported mobility as well as slower objective walking speed. ^{15 16 17} Furthermore, the association between DSL and walking speed was reported in an English cohort of older adults. ¹⁸ However, most prior studies have been cross-sectional. This longitudinal study assesses the association between DSL and walking speed using a population-based sample of older adults in the United States, the Health and Retirement Study (HRS).

Methods

Study Design and Participants

This study included data from the HRS, collected between 2006 and 2018. The HRS is a nationally representative longitudinal panel study of noninstitutionalized older adults in the United States. The HRS is sponsored by the National Institute on Aging (grant number NIA U01AG009740) and is conducted by the University of Michigan. Its protocol has been approved by the Institutional Review Board at the University of Michigan and all participants provided informed consent.

Participants in HRS are surveyed every 2 years. In 2006, half of HRS participants were randomly selected to receive a home visit and the other half received a phone interview. In 2008, the allocation was switched so that participants who had a home visit in 2006 received a phone interview and vice versa. This alternating interview allocation continued after 2008 such that each participant received a home visit every 4 years. Walking speed was assessed during the home visits in participants 65 years and older. ¹⁹

Measures

Walking Speed

The primary outcome was walking speed. It was measured at the participants' homes on a 98.5-inch course. Participants were instructed to walk at their normal pace. ¹⁹ Two timed trials were completed and their average was used to calculate walking speed in meters per second (m/s).

Sensory Loss

Visual functioning was assessed on a single item: “Is your eyesight excellent, very good, good, fair, or poor using glasses or corrective lens as usual?” Hearing functioning was similarly assessed on a single item: “Is your hearing excellent, very good, good, fair, or poor?” Participants who indicated using a hearing aid during the interview or any prior interview were asked to rate their hearing “using a hearing aid as usual.” VL and HL were defined as reporting fair or poor sensory ability on the respective question. Sensory loss was categorized as neither sensory loss (NSL), VL alone, HL alone, and DSL.

Covariates

Demographic covariates were self-reported in the interview and included age, sex, race (modeled as white vs. other), ethnicity (Hispanic vs. other), relationship status (married or partnered vs. other), and education (high school or less, some or full college, graduate degree).

Health covariates included body mass index (BMI) and chronic disease index. BMI was calculated based on measured height and weight. When either of these was not available, self-reported height or weight was used. Participants who had both measured and self-reported readings tended to underestimate their weight and overestimate their height. Missing BMI for a specific visit was imputed as the average BMI in the remaining visits. Chronic disease index was the number of self-reported diagnoses among diabetes, hypertension, heart disease, stroke, lung disease, non-skin cancer, and arthritis.

Statistical Analysis

Data

Vision and hearing were assessed at the baseline visit—defined for this study as the first home visit between 2006 and 2008. Analyses included the two subsequent home visits at years 4 and 8, and an additional visit at year 12 for participants with a baseline visit in 2006. Participants using walking aids at the baseline ( n  = 349) visit were excluded from the analysis. Legally blind participants and those with missing visual or hearing data at baseline ( n  = 116), as well as participants with missing BMI for all visits ( n  = 129), were also excluded.

Longitudinal Analysis

Mixed effects linear regression models were used to assess the association between sensory loss and walking speed. Interaction terms between sensory loss categories and years since baseline visit were included and tested for significance to determine if there was a difference in the rate of decline of walking speed by sensory loss category. Unadjusted and adjusted associations were reported. Adjusted models controlled for age (at baseline), sex, race, ethnicity, education, relationship status (modeled as time-varying covariate), BMI (time varying), and chronic disease index (time varying). Model marginal effects were used to calculate the predicted walking speed for each sensory loss category over time.

Time to Event Analysis

We assessed the association between sensory loss and incident slow walking speed (defined as walking speed slower than 0.6 m/s) ²⁰ and very slow walking speed (defined as walking speed slower than 0.4 m/s). The additional analysis for very slow walking speed was conducted because the analytic cohort composed of individuals 65 years and older, who have high prevalence of slow walking speed. In addition, a walking speed of 0.4 m/s or less has been shown to be associated with even greater mortality risk than a walking speed of 0.6 m/s as well as mobility limited only to within the household. ^{2 21} The analytic cohorts for these analyses are different from that for the longitudinal analysis, as individuals with the outcome at baseline were excluded from the analysis.

The time metric was years since baseline study visit. Participants who developed incident slow walking speed but had missing data for the previous visit were assumed to have developed the outcome at the midpoint between the current visit with slow walking speed and the previous visit with missing walking speed data. A similar method was used for incident very slow walking speed. Cox proportional hazard regression models were used to determine the relative hazard ratios (HRs) and 95% confidence intervals (CIs) for the association between sensory loss and incident slow and very slow walking speeds. The proportional hazard assumption was met in both analyses. Binned incidence analysis using Poisson's regression and discrete-time survival analysis using pooled logistic regression resulted in the same inference, and only the results from the Cox proportional regression are presented. The analyses adjusted for the same covariates listed earlier; however, relationship status, BMI, and chronic disease index were treated as time-fixed variables using their values from the baseline visit. All analyses were conducted using the R statistical software package version 3.5.3. ²²

Results

Study Population

This sample included 7,648 individuals: 947 reported VL alone, 1,197 reported HL alone, and 697 reported DSL. Participant characteristics are summarized in Table 1 . In general, individuals with DSL were older than individuals with NSL and more likely to be male, non-white, and Hispanic. Individuals with DSL were less likely to be married or partnered than individuals with NSL but more likely than individuals with VL alone. Education was generally higher in individuals with NSL, and chronic disease index was higher in individuals with DSL. A cross-tabulation of reported vision and hearing abilities is presented in Table 2 .

Table 1. Baseline characteristics of the study population.

	Overall N  = 7,648	NSL N  = 4,807	VL alone N  = 947	HL alone N  = 1 197	DSL N  = 697	p -Value
Percentage		62.9	12.4	15.7	9.1
Age, M ± SD	74.0 ± 6.7	73.4 ± 6.4	74.6 ± 6.9	75.1 + 7.0	75.7 ± 7.3	< 0.0001
Female	4,327 (56.6)	2,884 (60.0)	663 (70.0)	476 (39.8)	304 (43.6)	< 0.0001
Non-White race	1,074 (14.0)	649 (13.5)	198 (20.9)	103 (8.6)	124 (17.8)	< 0.0001
Hispanic ethnicity	568 (7.4)	259 (5.4)	108 (11.4)	80 (6.7)	121 (17.4)	< 0.0001
Married or partnered	4,860 (63.5)	3,097 (64.4)	523 (55.2)	812 (67.8)	428 (61.4)	< 0.0001
Education						< 0.0001
High school or less	5,872 (76.8)	3,553 (73.9)	791 (83.5)	923 (77.1)	605 (86.8)
College	1,153 (15.1)	798 (16.6)	109 (11.5)	177 (14.8)	69 (9.9)
Graduate	623 (8.1)	456 (9.5)	47 (5.0)	97 (8.1)	23 (3.3)
BMI, M ± SD	28.8 ± 5.7	28.8 ± 5.7	28.7 ± 5.9	28.6 ± 5.2	29.2 ± 5.8	0.25
Chronic disease index, M ± SD	2.2 ± 1.3	2.1 ± 1.2	2.4 ± 1.3	2.3 ± 1.3	2.6 ± 1.3	< 0.0001

Abbreviations: NSL, neither sensory loss; VL, vision loss; HL, hearing loss; DSL, dual-sensory loss; M, mean; SD, standard deviation.

Table 2. Cross-tabulation of baseline self-reported vision and hearing ability.

		Vision
		Excellent	Very good	Good	Fair	Poor
Hearing	Excellent	207 (2.7)	326 (4.3)	379 (5.0)	118 (1.5)	27 (0.4)
	Very good	164 (2.1)	780 (10.2)	731 (9.6)	183 (2.4)	65 (0.8)
	Good	166 (2.2)	615 (8.0)	1,439 (18.8)	452 (5.9)	102 (1.3)
	Fair	86 (1.1)	217 (2.8)	613 (8.0)	432 (5.6)	86 (1.1)
	Poor	31 (0.4)	59 (0.8)	191 (2.5)	113 (1.5)	66 (0.9)

Longitudinal Analysis

At the baseline visit, the average walking speed in the study population was 0.77 ± 0.25 m/s. Compared with individuals with NSL and after adjusting for all covariates, walking speed was slower by 0.05 m/s (95% CI = 0.04, 0.07), 0.02 m/s (95% CI = 0.003, 0.03), and 0.07 m/s (95% CI = 0.05, 0.08) in VL alone, HL alone, and DSL, respectively. Walking speed was reduced by 0.01 m/s (95% CI = 0.011, 0.012) per 1 year of older age at baseline. Therefore, in terms of their association with walking speed, VL was equivalent to being 4.4 years older, HL to being 1.4 years older, and DSL to being 5.7 years older. In the NSL group, walking speed declined by 0.014 m/s (95% CI = 0.013, 0.015) per year of follow-up. This rate of decline was not different among those reporting VL alone, HL alone, and DSL ( Table 3 ). Predicted walking speeds in the sensory groups after controlling for the covariates are shown in Fig. 1 .

Table 3. Longitudinal analysis—association of sensory loss with walking speed and rate of decline of walking speed.

Sensory loss	Number at baseline	Unadjusted β	Adjusted β
NSL	4,807	Reference	Reference
VL alone	947	−0.10 (−0.12, −0.09)	−0.05 (−0.07, −0.04)
HL alone	1,197	−0.03 (−0.04, −0.01)	−0.02 (−0.03, −0.003)
DSL	697	−0.12 (−0.14, −0.10)	−0.07 (−0.08, −0.05)
Year (change in walking speed over per year in the NSL group)		−0.015 (−0.016, −0.014)	−0.014 (−0.015, −0.013)
Sensory loss—year interaction
NSL		Reference	Reference
VL alone		0.0008 (−0.002, 0.003)	0.0006 (−0.002, 0.003)
HL alone		0.0000 (−0.002, 0.002)	−0.005 (−0.003, 0.002)
DSL		0.0022 (−0.001, 0.005)	0.0019 (−0.001, 0.005)

Figure 1.

Predicted walking speed over time in the different sensory groups.

Incident Slow and Very Slow Walking Speeds

Overall, 1,715 and 617 participants developed slow and very slow walking speeds during the follow-up period, yielding incidence rates of 5.3 and 1.6 per 100 person-years, respectively. The incidence of both slow and very slow walking speeds was higher in individuals with sensory loss than NSL individuals, with a large increase in incident very slow walking speed in DSL individuals (3.1 per 100 person-years) ( Table 4 , Fig. 2 , Fig. 3 ).

Table 4. Time to event analysis—incidence of slow and very slow walking speeds by sensory loss category.

	Incident slow walking speed (< 0.6 m/s)				Incident very slow walking speed (< 0.4 m/s)
Sensory loss	Number at baseline	PY	Events	Incidence a	Number at baseline	PY	Events	Incidence a
NSL	3,062	22,320	1,052	4.7	3,500	26,868	347	1.3
VL alone	463	3,110	232	7.5	592	4,276	87	2.0
HL alone	690	4,842	287	5.9	794	5,904	100	1.7
DSL	318	1,986	144	7.3	401	2,678	83	3.1
Overall	4,533	32,258	1,715	5.3	5,287	39,726	617	1.6

Abbreviations: DSL, dual-sensory loss; HL, hearing loss; NSL, neither sensory loss; PY, person-years; VL, vision loss.

^aIncidence per 100 person-years.

Figure 2.

Kaplan–Meier plot of survival without slow walking speed (walking speed < 0.6 m/s) in the different sensory groups.

Figure 3.

Kaplan–Meier plot of survival without very slow walking speed (walking speed < 0.4 m/s) in the different sensory groups.

Time to Event Analysis

After adjusting for the covariates, there was a greater risk of incident slow walking speed in each sensory loss category—adjusted HRs were 1.43 (95% CI = 1.25, 1.65), 1.29 (95% CI = 1.13, 1.48), and 1.35 (95% CI = 1.13, 1.61) in VL alone, HL alone, and DSL groups, respectively. The risk of slow walking speed was not different in DSL individuals compared with individuals with either VL or HL alone ( Table 5 ).

Table 5. Time to event Cox proportional hazard model analysis—association between sensory loss and incident slow and very slow walking speeds.

	Slow walking speed (< 0.6 m/s)		Very slow walking speed (< 0.4 m/s)
Sensory loss	Unadjusted HR (95% CI)	Adjusted HR (95% CI)	Unadjusted HR (95% CI)	Adjusted HR (95% CI)
NSL	Ref.	Ref.	Ref.	Ref.
VL alone	1.71 (1.48, 1.97)	1.43 (1.25, 1.65)	1.63 (1.29, 2.06)	1.21 (0.96, 1.54)
HL alone	1.30 (1.14, 1.48)	1.29 (1.13, 1.48)	1.33 (1.07, 1.67)	1.30 (1.03, 1.63)
DSL	1.68 (1.41, 2.00)	1.35 a (1.13, 1.61)	2.58 (2.03, 3.28)	1.89 b (1.47, 2.43)

Abbreviations: CI, confidence interval; DSL, dual-sensory loss; HL, hearing loss; HR, hazard ratio; NSL, neither sensory loss; VL, vision loss.

^aDSL vs. VL alone HR = 0.94 (0.76, 1.16); DSL vs. HL alone HR = 1.04 (0.85, 1.28).

^bDSL vs. VL alone HR = 1.56 (1.14, 2.12); DSL vs. HL alone HR = 1.46 (1.09, 1.96).

There was similarly a greater risk of incident very slow walking speed in HL alone and DSL groups after adjusting for the covariates—adjusted HRs were 1.30 (95% CI = 1.03, 1.63) and 1.89 (95% CI = 1.47, 2.43) in HL alone and DSL, respectively. For participants with VL alone, the risk was increased (HR = 1.21, 95% CI = 0.96, 1.54) without reaching statistical significance. The risk of incident very slow walking speed was significantly higher in individuals with DSL compared with individuals with either VL or HL alone ( Table 5 ).

Discussion

In this study, we found that older adults who reported VL and/or HL walked slower than their NSL counterparts. The differences in walking speed by sensory loss continued throughout a follow-up period of up to 12 years. Furthermore, older adults with reported sensory loss were at increased risk of developing slow walking speed during follow-up. Older adults with DSL were particularly at risk of developing very slow walking speed, which is a significant risk factor for functional dependence and mortality.

In this study, we showed that older adults with sensory loss walked more slowly than their counterparts without sensory loss. Mechanistically, individuals with sensory loss may slow their gait in the short term because of the reduced awareness of the environment, change in gait, and fear of falling that occur as a consequence of their sensory loss. ^{23 24 25 26} Over time, this adaptation may translate to a biological change and an inherently slower gait speed. In addition, sensory loss can impact the general health of the individual through social isolation, ^{27 28 29} decreased cognition, ⁷ and depressed mood. ^{30 31} This adverse impact on health would then be reflected in the individual's walking speed. It is possible then that addressing sensory loss and teaching compensatory strategies may mitigate at least some of the negative consequences of sensory loss on reduced walking speed. An important question then arises from this discussion: is there an age or a walking speed threshold beyond which addressing sensory loss becomes less impactful?

The answer to the question is likely to be “yes,” and this is supported by our finding that the decline in walking speed is similar across the different sensory loss groups. This similar rate of decline indicates that the relationship has already been established to create the differences in walking speeds at baseline; and the impact of sensory loss on the decline in walking speed has largely ceased in the majority of the study population, who are all older than 65 years and have slow gait speeds—average of 0.77 m/s at baseline. A similar trajectory of walking speed decline across HL levels has been documented in a prior study of a similarly aged population. ³² Studies using the Salisbury Eye Evaluation Study found that VL is associated with consistently slower walking speed, but older adults with VL did not have a more rapid decline in walking speed. This result may indicate that older adults with VL may develop compensatory strategies that compensate for and mitigate against continued decline in gait speed over time or that there is “floor” effect to which VL negatively impacts gait speed but only to a certain level. ^{33 34} Perhaps rather than an age limit, there is a walking speed threshold for intervention in these older adults. This idea is supported by our survival analysis, which excludes individuals with slow speeds at baseline. In this analysis, we found that older adults with sensory loss were at increased risk of incident slow and very slow walking. Therefore, addressing sensory loss before they reach these levels of slow speed may mitigate the risks of reaching these slow speeds and their adverse health consequences.

Furthermore, the combined effect of VL and HL resulted in even greater incidence than either VL or HL alone of very slow walking speed. However, their combined effect was not greater than either VL or HL alone on the incidence of slow walking. That is, vision and hearing loss act synergistically to reduce walking speed; however, this synergy may occur only at very slow speeds, where older adults are at high risk of functional dependence. ³⁵ One approach older adults use to adapt to sensory loss is sensory substitution—using the functional sensory modality to compensate for the one with reduced function. ³⁶ Older adults with DSL would be less able to compensate when both vision and hearing are lost, and it appears that this inability to compensate may occur at very slow walking speed. Stepping becomes predominantly intermittent rather than continuous at very slow walking (< 0.4 m/s), ³⁷ and this gait characteristic may be associated with the onset of synergy of the effects of VL and HL.

The strengths of this study include using data from a prospective population-based cohort and its longitudinal analytic approach, adding value to the existing literature that has mostly been cross-sectional. Furthermore, the time to event analysis revealed an important temporal relationship between sensory loss and incident slow and very slow walking speeds, with synergistic combined effects of VL and HL on very slow walking. One potential limitation of this study is its use of self-reported measures of sensory function; however, this may also be viewed as a strength, as measuring vision using a single objective measure may not capture the full dimensionality of the effect of vision on function. ³⁸ Another limitation is the older age of the population and the fact that a majority of participants walked slowly at baseline; future studies should examine the association between sensory function and walking speed in a younger cohort, where a difference in the rate of decline of walking speed by sensory function may be more apparent.

In summary, older adults with reported sensory loss walked slower and are at increased risk of developing incident slow walking speed than older adults without sensory loss, and those with DSL may be at a particularly high-risk group for slow walking. Furthermore, the impact of VL on walking speed was equivalent to being 4.4 years older, HL to being 1.4 years older, and DSL to being 5.7 years older, highlighting the magnitude of the impact of sensory loss on walking speed. Overall, these results indicate the negative impact of sensory loss on walking speed among older adults.